[time-nuts] Pre-averaging Phase Data

[Poul-Henning Kamp]

--------
In message <4D24832F-8127-411A-8566-973AC6F5D473 at n1k.org>, Bob kb8tq writes:

>In general the way you go from 1 to 10 to 100 seconds is to decimate
>( = throw away) the data to get a phase record at the tau you are after.

The problem with that, is that decimation may _also_ throw away noise.

As a general scientific rule, if you measured, you use the measurements,
you cannot "pick & choose" amongst your measurements, not even if
you do it in a certifiably random way.

As I understand it, that is one of the minor problems the Modified
Allan Variance was (also) supposed to fix ?

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.